If tidying, clear unchanged records.

Anything that hasn’t been inserted or updated will have an ‘old’ timestamp - the record will be from a deleted model instance that wasn’t cleaned up correctly.

lib/thinking_sphinx/real_time/populator.rb

@@ -7,6 +7,7 @@ def self.populate(index)
 
   def initialize(index)
     @index = index
+    @started_at = Time.current
   end
 
   def populate
@@ -17,12 +18,14 @@ def populate
       instrument 'populated', :instances => instances
     end
 
+    transcriber.clear_before(started_at) if configuration.settings["real_time_tidy"]
+
     instrument 'finish_populating'
   end
 
   private
 
-  attr_reader :index
+  attr_reader :index, :started_at
 
   delegate :controller, :batch_size, :to => :configuration
   delegate :scope,                   :to => :index

lib/thinking_sphinx/real_time/transcriber.rb

@@ -5,6 +5,12 @@ def initialize(index)
     @index = index
   end
 
+  def clear_before(time)
+    execute <<~SQL.strip
+      DELETE FROM #{@index.name} WHERE sphinx_updated_at < #{time.to_i}
+    SQL
+  end
+
   def copy(*instances)
     items = instances.select { |instance|
       instance.persisted? && copy?(instance)